betaFeedback

WorksPhysics and Astronomy

Topological Data Analysis of Ion Migration Mechanism

Abstract

Topological data analysis based on persistent homology has been applied to the molecular dynamics simulation for the fast ion-conducting phase (αα-phase) of AgI, to show its effectiveness on the ion-migration mechanism analysis.Time-averaged persistence diagrams of αα-AgI, which quantitatively records the shape and size of the ring structures in the given atomic configurations, clearly showed the emergence of the four-membered rings formed by two Ag and two I ions at high temperatures. They were identified as common structures during the Ag ion migration. The averaged potential energy change due to the deformation of four-membered ring during Ag migration agrees well with the activation energy calculated from the conductivity Arrhenius plot. The concerted motion of two Ag ions via the four-membered ring was also successfully extracted from molecular dynamics simulations by our approach, providing the new insight into the specific mechanism of the concerted motion.

Introduction

To realize all-solid-state ion batteries, ion-conducting electrolytes with high performance have been developed so far[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]. These materials, called superionic conductors, tend to exhibit complex ion migration mechanisms, such as the ion migration with concerted motion[13, 14, 15, 16], utilizing rotating anion[1, 4, 6, 7, 9, 10, 11, 12, 16, 17], and the ion migration in glass materials[8], which deviates from a single-ion hopping in a uniform lattice. Since atoms around mobile ions are arranged in a wide variety of configurations in such materials, extracting representative ion migration mechanisms and rate-limiting step is not easy. Even for bulk materials such as Li7La3Zr2O12\mathrm{Li}_7\mathrm{La}_3\mathrm{Zr}_2\mathrm{O}_{12}, the activation energy for each Li ion migration during molecular dynamics (MD) simulations is widely distributed from 0 to 0.6 eV[18]. Therefore, it is challenging to construct a guideline for material design based on the migration mechanism.

Topological data analysis (TDA) based on persistent homology has been used for structural analysis of amorphous materials[19, 20, 21, 22], glasses[23, 24, 25, 26, 27], porous materials[28], and complex molecular liquids[29]. This analysis can characterize point cloud data (ex. atomic structures), by projecting them onto a two-dimensional density distribution of the rings and pores in the system. Extracting representative ring/pore structures from the distribution and performing inverse analysis have contributed to the elucidation of the hidden order in the complex materials. In recent years, persistence diagrams, the distribution of ring and pore structures, have been combined with machine learning and other techniques to predict physical properties such as thermal conductivity properties[30, 31], stress-strain curves[32], formation enthalpy[33], and so on[34].

TDA was also used for dynamic phenomena. For example, it was applied for glass formation[26] and metal nucleation processes[35]. Based on defect chemistry, the ion migration is a phenomenon that occurs around ion vacancies. By considering vacancies in ionic conductors as large voids inside rings and pores in the persistence diagram, ion migration can be analyzed using TDA. By performing the inverse analysis along the ring/pore structures with high density, a representative ion migration mechanism within the material is considered to be extracted from the persistence diagram. Since the obtained rings/pores are composed of multiple ions, it is possible to discuss ion migration as a many-body problem.

In this study, we analyzed the Ag migration mechanism of α\alpha-AgI[36, 37, 38, 39, 40, 41, 42, 43] [44], a classical superionic conductor, as an example to elucidate complex migration mechanisms such as concerted motion. To our best knowledge, it is the first report to apply topological data analysis based on persistent homology to the ion migration study. The advantage of this method over existing ones is that it does not require any prior information like collective variables for constrained MD simulations and ion migration paths based on the crystallography, once the sufficient MD simulation trajectory is obtained. Note that the potential energy surface of the whole system mainly used in transition state search does not necessarily work for analyzing migration paths in these ionic conductors, because multiple ion migrations could occur simultaneously in a simulation cell. In addition, in the highly-disordered system, the activation energy obtained from nudged elastic band (NEB) method45 is strongly affected by the position of mobile ions, because of the electrostatic interaction between them. This suggests that plenty of NEB calculations with different mobile-ion configurations are needed to obtain a reasonable ion migration mechanism. Therefore, this work becomes a milestone to understand the ion migration mechanism for highly-disordered system like glass materials[8] and compounds with complex-anions[1, 4, 6, 7, 9, 10, 11, 12, 16, 17].

Method

Molecular dynamics simulations

NVT-MD simulations of α\alpha-AgI were performed using MD calculation package (Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS))46 (version 29 October 2020)). Note that we chose α\alpha-AgI as a model system to demonstrate the effectiveness of TDA for the ion migration mechanism for the following reasons; one is its high ionic conductivity[37] [47] making it easy to sample the ion migration during MD simulation. another is that the classical potential48,49 is available in this system, reducing the calculation cost. The other reason is that α\alpha-AgI shows concerted motion of Ag ions[41, 42], although it is consisted of ordered matrix, the bcc lattice of I ions. Therefore, it is easy to verify whether the ion migration mechanism obtained from TDA is reasonable or not, by comparing it with that suggested from the analysis and discussion based on the crystal structure.

The detail of the calculation condition is followings. Nosé–Hoover thermostat and barostat51 was employed for the temperature control in all MD simulations. Periodic boundary conditions were applied along all the Cartesian directions. Newton’s motion equations were integrated using the customary Verlet’s algorithm with a time-step length of 0.005 ps for MD simulations. 3-ns MD simulations with a cubic cell, 20.42 ×\times 20.42 ×\times 20.42 Å, consisting of 256 Ag and I ions (4×\times4×\times4 supercell of α\alpha-AgI (Figure 1)) were performed at various temperatures from 360 to 1000K. Previous study44 have reported that the finite size effect is negligibly small in AgI system, in terms of diffusion coefficients calculation. This suggests that our selected cell size does not affect the ion migration process during these MD simulations. We used the Parrinello-Rahman-Vashishta potential48,49, for the interatomic potential for α\alpha-AgI. It is known that Parrinello-Rahman-Vashishta potential represents Ag-I and I-I interactions well and can reproduce many physical properties such as diffusion coefficients48 and phase transitions49. However, it has been reported that this potential cannot fully reproduce Ag-Ag interactions due to the lack of treatment for covalent interactions between Ag ions[39] and atomic polarization[40]. Nevertheless, the accuracy of the potential is not a matter of concern in this study, since our goal is to obtain the essential processes hidden in the complex ion migration dynamics containing concerted motions. Also note that NVT-MD simulations are intentionally conducted to maintain α\alpha-AgI structure at low temperatures (360–440K). This makes us easier to compare the persistence diagram obtained at different temperatures since the matrix, I bcc lattice, and cell parameter is constant.

Crystal structures of alpha-AgI and conductivity Arrhenius plot

Figure 1. (a) Crystal structure of α\alpha-AgI (white: Ag black: Ag vacancy, purple: I). I ions show a bcc structure, while Ag sites are located at the vertexes of the Voronoi polyhedron (black bold lines in the figure), which is composed of the perpendicular bisectors of each I ion. Ag ions occupy two of the twelve sites (12d sites). (b) An example of the structure of α\alpha-AgI when Ag ions are specifically placed in two of these sites. (c) Conductivity Arrhenius plot of α\alpha-AgI during classical MD simulations50. The dashed curve represents the fitting result of conductivity above 440K using Arrhenius equation, ln⁡σT=−EA/RT+A\ln\sigma T=-E_A/RT+A.

To calculate time-averaged potential energy for ring-structures obtained from TDA, we also conducted constrained MD simulations. In these calculations, smaller cubic cell, 10.21 ×\times 10.21 ×\times 10.21 Å, consisting of 32 Ag and I ions (2×\times2×\times2 supercell) was employed. In the simulation cell, we introduced one four-membered ring consisting of two Ag and two I ions obtained from TDA. Ions in this ring structure have been fixed during MD simulations by setting their force and velocity to zero. We prepared 16 simulation cells

with different four-membered rings along the distribution obtained from the persistence diagram. Then, 4-ns MD simulations of these cells were conducted at 1000K. The difference of time-averaged potential energy between these simulations should be originated from the shape of four-membered ring and its surrounding environment, since the sampling time was long enough and ions except that in fixed four-membered ring could move freely. Therefore, we can discuss the change in potential energy during Ag ion migration via the four-membered ring based on time-averaged potential energy of these MD simulations.

Topological data analysis

Persistent homology is a mathematical framework for analyzing n-dimensional holes in the given discrete data such as atomic coordinates. A software package, Homcloud47,52, is employed for TDA based on persistent homology. Here, we only show a brief outline of its concept. See also reference 19 and 27 for the theoretical details and the application to materials. We will focus on one-dimensional holes (i.e. ring structures) hereafter. Note that one-dimensional hole in homology is defined as the states, which has a one-dimensional boundary (i.e., a space bounded by lines connecting atoms) but no two-dimensional boundary (i.e., the surface created by connecting atoms is not entirely filled by a sphere of radius rr, assuming there is the sphere with a certain radius rr at each atomic coordinate). Figure 2 is a schematic diagram of how rings are analyzed in TDA based on persistent homology. First, a sphere with the common radius, rr, is placed at each of atomic coordinates regardless of its element. By increasing rr, new edges are spanned one by one when two spheres are overlapped. At a particular radius (r=rbr=r_{\mathrm{b}}), some atoms are connected to each other and form a ring (top right figure in Figure 2). In this paper, we name this diameter 2rb2r_{\mathrm{b}} as the birth diameter and assign it to x-coordinate on the persistence diagram. At larger radius (r=rdr=r_{\mathrm{d}}), this ring structure is filled by the sphere as shown in the bottom left figure. Here, the diameter 2rd2r_{\mathrm{d}} is named as death diameter and assigned to y-coordinate. Thus, a persis-

Schematic diagram of how ring structures in an atomic configuration are analyzed by topological data analysis

Figure 2. Schematic diagram of how ring structures in an atomic configuration are analyzed by topological data analysis (TDA) based on persistent homology. In this analysis, we place a sphere with same radius, rr, at each atomic coordinate in the system. By increasing rr, some atoms are connected to each other to form the ring. The diameter of the sphere when the ring is formed (2rB2r_{\mathrm{B}}) is assigned to the x-coordinate as birth diameter. At larger radius (r=rDr=r_{\mathrm{D}}), this ring structure is filled by the spheres. Here, the diameter 2rD2r_{\mathrm{D}} is named as death diameter and assigned to y-coordinate.

tence diagram is obtained as a two-dimensional scatter plot of birth-death pairs. To make it easy to understand visually, We limit the area to the square of [0.0,12.0]×[0.0,12.0][0.0, 12.0]\times[0.0, 12.0] (A˚2\text{\AA}^{2}) and divided it into 200×200200\times200 meshes. We evaluated the number of birth-death pairs in each mesh as log⁡10P(b,d)\log_{10}P(b,d).

Results and Discussion

Change in the persistence diagram against ionic conductivity in α\alpha-AgI

α\alpha-AgI consists of a bcc I lattice with twelve stable Ag sites (12d sites), as shown in Figure 1(a). These stable Ag sites correspond to the vertex of the Voronoi polyhedron for the bcc lattice, the polyhedron created by the perpendicular bisectors between I ions. Ag ions only occupy two of these stable sites like the structure in fig. 1(b). Therefore, once it is thermally activated, Ag ion can utilize unoccupied vacancy sites for ion migration. According to previous studies[36, 37], six out of seven Ag ions migrate along the sides of the Voronoi polyhedron, the black line in the figure. On the other hand, the rest of Ag ions is transported through the face-center position (0.5 0.5 0) in the unit cell. Figure 1(c) is a conductivity Arrhenius plot of α\alpha-AgI during MD simulations at various temperatures. We used Green-Kubo formula for the conductivity calculation50,53–57. Due to the poor convergence of these calculations, the conductivity calculations used the results of MD simulations extended to 25 ns. α\alpha-AgI showed superionic conduction at temperatures higher than 440 K. The obtained conductivity above 440 K is in good agreement with the experimental value (≈100\approx10^{0} (S/cm))[37].

The change in the local structure of α\alpha-AgI during Ag migration has been confirmed prominently as that in ring structures in the persistence diagram. Figure 3 shows time-averaged persistence diagrams during MD simulations at various temperatures. We conducted TDA for the atomic configurations every 0.05 ps during MD simulations and obtained the averaged persistence diagram of 6,000 configurations. Two types of ring structures are mainly distributed around (Birth, Death)=(b,d)=(2.8,4.5) and (4.2,4.4) in the persistence diagrams for MD simulations at 360K and 400K, where ionic conductivity is low. On the other hand, the ring-structure distribution along the black dashed line increases on the persistence diagrams at 550K and 1000K. Therefore, these ring structures could correspond to the ones for Ag migration. Note that the same change in the ring structure has been confirmed in the persistent diagrams obtained from the NPT-MD simulations, taking into account the phase transition to β\beta-AgI (see section S1 in the supporting information).

Time-averaged persistence diagrams of the ring structure during MD simulations of α-AgI at several temperatures

Figure 3. Time-averaged persistence diagram of the ring structure during MD simulations of α-AgI for several temperatures. This persistence diagram represents a probability density (PP) distribution normalized so that the integrated value over the squared regions [0.0, 12.0] × [0.0, 12.0] (Å2^2) becomes unity. The color in the figure is scaled according to log⁡10P\log_{10}P and represents the probability density per the mesh of 0.05 × 0.05 Å2^2.

Inverse analysis for the persistence diagram of α\alpha-AgI

From the inverse analysis52 of the persistence diagram, a model of Ag ion migration via four-membered rings has been obtained. Here, the inverse analysis means to calculate averaged property for each mesh in the persistence diagram, recorded during the TDA. Inverse analysis of the persistence diagram at 1000K confirmed that these ring structures form a four-membered ring consisting of two Ag and I ions (Fig. S2). The persistence diagram of the four-membered rings in the system and the results of their inverse analysis (the average I-I, Ag-I, and Ag-Ag distances in the four-membered rings at each mesh) are summarized in Figure 4(a). As shown in the figure, the distance between I ions is almost constant. On the other hand, the length of longest Ag-I bond in the four-membered ring decreases as birth decreases. Similarly, the distance between Ag ions was also shortening with decreasing death. These results show that the four-membered ring deforms, as the birth decreases along the dashed black line in the persistence diagram. Figure 4(b) shows the structure of the four-membered ring reproduced based on these average distances. Note that I ions other than that in the 4-membered ring occupy the ideal positions of bcc structure in Figure 4(b). The black line is the Voronoi polyhedron of an ideal I bcc lattice (≈Ag migration path). Ag1 moved from one vertex to the other in the Voronoi polyhedron, indicating that Ag hopping migration has been reproduced by the deformation of the four-membered ring obtained from the persistence diagram.

Persistence diagrams and structures of four-membered rings

Figure 4. (a) Persistence diagram only showing four-membered rings and the distribution of averaged I-I, Ag-I, and Ag-Ag distances in each (b,d) for these four-membered rings in the persistence diagram. (Calculated for each 0.05 × 0.05 Å2^2 meshes.) The color in top left figure is scaled according to log⁡10P\log_{10}P and represents the probability density per the mesh of 0.05 × 0.05 Å2^2. On the other hand, that in other figures is scaled according to the distance (Å) between ions. (b) Ag migration obtained by the deformation of four-membered ring along the black dashed line on the persistence diagram. Note that I ions except that in the four-membered ring show ideal I bcc lattice positions. (white: Ag, purple: I)

In these four-membered rings, the interaction between Ag1 and Ag2 ions via the I-lattice has been confirmed from bond valence analysis (Fig. S3). Such a picture of Ag migration via interaction with I-lattice has been claimed in previous studies. For example, covalent-bond formation between nearest neighbour Ag and I ions has been reported from the bond-overlap population analysis of ab initio MD simulation for molten AgI[39]. In addition, ab initio MD simulation and THz-

frequency Raman polarization-orientation measurement suggest that anharmonic vibrations in the I lattice are important for Ag migration[43].

Fig. 5 and the video in Supporting information summarize possible Ag migrations via the four-membered ring. During Ag migration in Figure 5(a)–(Figure 5c), the four-membered ring deforms along the dashed black line in Figure 4(a). On the other hand, the four-membered ring shape is maintained in the latter part of Ag migration in Figure 5(d). There are four possible migrations. The first one (Figure 5(a)) shows Ag hopping and subsequent returning to its original site. In the second one (Figure 5(b)), Ag1 further migrates to the other site. In this reaction, Ag2 returns to the almost same position. Therefore, this migration is considered to be the single Ag1 hopping. The third one (Figure 5(c)) is a concerted motion, where Ag1 migrates from one site to the other and Ag2 hopping occurs immediately after Ag1 migration. Figure 5(d) also shows a concerted migration of Ag ions. However, it occurs without the deformation of the four-membered ring. As shown in the figure, Ag ions migrate along the hexagon on the Voronoi polyhedron of I bcc, maintaining the shape of the four-membered ring.

Ag migration mechanisms

Figure 5. Ag migration obtained by repeating the deformation of the four-membered ring as shown in Figure 4(b) (i.e., high (b,d) → low (b,d) → high (b,d)); (a) backward Ag migration, (b) single Ag hopping, (c) concerted motion of two Ag ions, and (d) concerted motion of Ag ions maintaining the four-membered ring shape. The figure is created with (111) as the normal vector, and (b,d) represents the values of birth and death of the four-membered ring. (white: Ag, purple: I)

From the analysis of Ag hopping directions between nearest neighbour sites, the ratio of these four Ag migrations are estimated to be around (a) 42%, (b) 19%, (c) 26%, and (d) 13%, respectively (Fig. S4). This suggests that concerted motions such as (c) and (d) occur with about 39% probability. The existence of concerted motion in AgI and similar class of material (Ag2Se) has been claimed based on the Haven ratio for ionic conductivity[42] and the velocity correlation function of Ag ions during MD simulations[38]. These results are consistent with the obtained concerted motion in Figure 5(c) and (d). The advantage of TDA from conventional studies is that we can extract the concrete model of concerted ion migration as shown in Figure 5.

The Ag migration via the four-membered ring obtained from the persistence diagram corresponds to the representative Ag migration in α\alpha-AgI. Figure 6 shows the time-averaged potential energy during 4-ns MD simulations with four-membered rings obtained from the persistence diagram fixed in a 2×2×22\times2\times2 α\alpha-AgI supercell. Since ions other than that in the four-membered ring moved freely and the sampling time was long enough, the energy difference in these MD simulations originated from the four-membered ring shape. Therefore, fig. 6 shows the energy profile for the deformation of the four-membered ring along the black dashed line in the persistence diagram in Figure 4(a). The averaged potential energy change due to the deformation of four-membered ring during Ag migration is about 0.12 eV. Note that the energy profile for three types of Ag migration in fig. 5(a)-(c) should be the same as that of Figure 6. This is because these Ag migrations can be reproduced by a point-symmetric operation of the four-membered ring on the black dashed line in Figure 4(a) using the middle point of two I ions in the ring. The averaged potential energy change due to the deformation of four-membered ring, which can reproduce Ag migration, is in good agreement with the activation energy obtained from the conductivity Arrhenius plot. This shows that the persistence diagram is able to extract a local but representative structure for Ag migration.

Time-averaged energy during MD simulations

Figure 6. Time-averaged energy during 4-ns MD simulations for 2×2×22\times2\times2 α\alpha-AgI supercells with one fixed four-membered ring obtained from a persistence diagram along the black dashed line in Figure 4(a).

The comparison between independent and four-membered ring Ag migration

Independent migration of each Ag ion like a random-walk model58 does not form the four-membered rings specific to the reproduced Ag migration, showing that the interaction between ions selectively creates them during MD simulations. Figure 7(left) shows a persistence diagram of an ideal I bcc lattice with Ag ions randomly placed on the migration path (i.e. the edges and vertexes of the Voronoi polyhedron for I bcc lattice). Since Ag ions are placed almost randomly along the migration path59, the resulting persistence diagram is considered to be equivalent to that for the random-walk model. We created 6000 random configurations of 4×4×44 \times4 \times4 supercells and calculated their averaged persistence diagram so that the calculation condition for it matches that for the persistence diagram for MD simulations. The black open squares in the figure represent the (b,d) positions of rings obtained when Ag ions are only placed at the stable site (the vertexes of the Voronoi polyhedron). The distribution among these black open squares was found to increase, when Ag ions are randomly placed on the migration path in the I bcc lattice. However, ring structures along the black dashed line observed in MD simulations at 550 and 1000K were hardly obtained in the persistence diagram of I bcc lattice with random Ag configurations. Therefore, we conclude that independent migration of Ag ions cannot explain the ones during MD simulations and the concerted motion of Ag ions induced by the interaction between them actually contributes to Ag ion conductivity in α\alpha-AgI.

Persistence diagrams of random Ag positions and MD at 550 K

Figure 7. (left) Persistence diagram of I bcc lattice with Ag ions randomly placed on the migration path, (edges and vertexes of the Voronoi polyhedron in Figure 1(a)). (right) Time-averaged persistence diagram during MD simulation of α\alpha-AgI at 550 K. The black open squares in the figure represent the (b,d) positions of the ring structure obtained from I bcc lattice with Ag ions randomly placed at the stable (12d) site (edges of the Voronoi polyhedron).The color in the figure is scaled according to log⁡10P\log_{10}P and represents the probability density per the mesh of 0.05×0.05 A˚20.05 \times0.05\ \text{\AA}^{2}.

Conclusion

To reveal the ion migration mechanism in the superionic conductor, we performed classical molecular dynamics simulations for α\alpha-AgI and made a series of topological data analysis based on persistent homology. When α\alpha-AgI shows superionic conductivity, characteristic Ag-I ring structures were newly detected on the persistence diagram. They were identified as four-membered rings formed by two Ag and I ions, which are common to the Ag ion migration. The averaged potential energy change due to the deformation of four-membered ring during Ag migration is about 0.12 eV, which is in good agreement with the activation energy obtained from the conductivity Arrhenius plot. Therefore, we conclude that the Ag ion migration via these four-membered rings is the representative Ag migration process in α\alpha-AgI. In this migration, two Ag ions showed concerted motion and the ratio of such Ag ion migration was about 39%. On the other hand, the number of the corresponding four-membered rings became significantly small, if random-walk of Ag ions in I bcc lattice is assumed. Therefore, based on topological data analysis of the MD trajectories, we conclude that it is successfully elucidated that Ag ions migrate not independently but concertedly via the four-membered ring in α\alpha-AgI.

Supplementary Material

See the supplementary material for the persistence diagram of AgI during 3-ns NPT-MD simulations (section S1), the distribution of most frequent number of atoms in the ring in persistence diagram during the MD simulation at 1000K (section S2), bond valence analysis of four-membered rings obtained from inverse analysis of persistence diagram (section S3), The result of the direction analysis of single Ag hopping during MD simulation at 1000K (section S4), and videos for Ag migrations through the deformation of four-membered ring obtained from the topological data analysis based on persistent homology.

Acknowledgments

This work was supported by JSPS KAKENHI Grant-in-Aid for Scientific Research on Innovative Areas “Hydrogenomics”, No. JP18H05513 and Grant-in-Aid for Challenging Research (Pioneering, No. 19H05514). The computation in this work has been done using the facilities of the Supercomputer Center, the Institute for Solid State Physics, the University of Tokyo.

Author Declarations

Conflict of Interest

The authors have no conflicts to disclose.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. Note that the input files for MD simulations, representative output files, and the calculation codes used in this paper are available on Materials Cloud platform60. See DOI: https://doi.org/10.24435/materialscloud:zn-5t

Reference

References

  1. [1]S. Orimo, Y. Nakamori, J. R. Eliseo, A. Züttel, and C. M. Jensen, “Complex hydrides for hydrogen storage,” Chem. Rev. 107, 4111–4132 (2007).
  2. [2]R. Murugan, V. Thangadurai, and W. Weppner, “Fast lithium ion conduction in garnet-type Li₇La₃Zr₂O₁₂,” Angew. Chem. Int. Ed. 46, 7778–7781 (2007).
  3. [3]N. Kamaya, K. Homma, Y. Yamakawa, M. Hirayama, R. Kanno, M. Yonemura, T. Kamiyama, Y. Kato, S. Hama, K. Kawamoto, and A. Mitsui, “A lithium superionic conductor,” Nat. Mater. 10, 682–686 (2011).
  4. [4]M. Matsuo and S. Orimo, “Lithium fast-ionic conduction in complex hydrides: Review and prospects,” Adv. Energy Mater. 1, 161–172 (2011).DOI
  5. [5]Y. Seino, T. Ota, K. Takada, A. Hayashi, and M. Tatsumisago, “A sulphide lithium super ion conductor is superior to liquid ion conductors for use in rechargeable batteries,” Energy Environ. Sci. 7, 627–631 (2014).DOI
  6. [6]A. Unemoto, M. Matsuo, and S. Orimo, “Complex hydrides for electrochemical energy storage,” Adv. Funct. Mater. 24, 2267–2279 (2014).DOI
  7. [7]R. Mohtadi and S. Orimo, “The renaissance of hydrides as energy materials,” Nat. Rev. Mater. 2, 16091 (2016).DOI
  8. [8]K. Ohara, A. Mitsui, M. Mori, Y. Onodera, S. Shiotani, Y. Koyama, Y. Orikasa, M. Murakami, K. Shimoda, K. Mori, T. Fukunaga, H. Arai, Y. Uchimoto, and O. Za, “Structural and electronic features of binary Li₂S-P₂S₅ glasses,” Sci. Rep. 6, 21302 (2016).DOI
  9. [9]S. Kim, H. Oguchi, N. Toyama, T. Sato, S. Takagi, T. Otomo, D. Arunkumar, N. Kuwata, J. Kawamura, and S. Orimo, “A complex hydride lithium superionic conductor for high-energy-density all-solid-state lithium metal batteries,” Nat. Commun. 10, 1081 (2019).DOI
  10. [10]K. Sau, T. Ikeshoji, S. Kim, S. Takagi, K. Akagi, and S.-i. Orimo, “Reorientational motion and Li⁺-ion transport in Li₂B₁₂H₁₂ system: Molecular dynamics study,” Phys. Rev. Materials 3, 075402 (2019).DOI
  11. [11]K. Kisu, K. Sim, T. Shinohara, K. Zhao, A. Züttel, and S. Orimo, “Monocarborane cluster as a stable fluorine-free calcium battery electrolyte,” Sci. Rep. 11, 7563 (2021).DOI
  12. [12]K. Sau and T. Ikeshoji, “Ring mechanism of fast Na⁺ ion transport in Na₂LiFeTeO₆: Insight from molecular dynamics simulation,” Phys. Rev. Materials 6, 045406 (2022).DOI
  13. [13]The migration of multiple ions triggered by that of one ion. Here, the term, concerted motion, is defined in a broad sense. In this article, if other ions move in conjunction with the migration of one ion, it is considered to be a concerted motion, regardless of the direction of transport, etc.
  14. [14]X. He, Y. Zhu, and Y. Mo, “Origin of fast ion diffusion in super-ionic conductors,” Nat. Commun. 8, 15893 (2017).DOI
  15. [15]Z. Zhang, Z. Zou, K. Kaup, R. Xiao, S. Shi, M. Avdeev, Y.-S. Hu, D. Wang, B. He, X. Li H., Huang, L. F. Nazar, and L. Chen, “Correlated migration invokes higher Na⁺-ion conductivity in NaSICON-type solid electrolytes,” Adv. Energy Mater. 9, 1902373 (2019).DOI
  16. [16]K. Sau, T. Ikeshoji, S. Kim, S. Takagi, and S. Orimo, “Comparative molecular dynamics study of the roles of anion-cation and cation-cation correlation in cation diffusion in Li₂B₁₂H₁₂ and LiCB₁₁H₁₂,” Chem. Mater. 33, 2357–2369 (2021).DOI
  17. [17]Z. Zhang and L. F. Nazar, “Exploiting the paddle-wheel mechanism for the design of fast ion conductors,” Nat. Rev. Mater. 7, 389–405 (2022).DOI
  18. [18]B. C. Wood, J. B. Varley, K. E. Kweon, P. Shea, A. T. Hall, A. Grieder, M. Ward, V. P. Aguirre, D. Rigling, V. E. Lopez, C. Stancill, and A. N., “Paradigms of frustration in superionic solid electrolytes,” Phil. Trans. R. Soc. A 379, 20190467–20190467 (2021).DOI
  19. [19]Y. Hiraoka, T. Nakamura, A. Hirata, E. G. Escolar, K. Matsue, and Y. Nishiura, “Hierarchical structures of amorphous solids characterized by persistent homology,” Proc. Natl. Acad. Sci. USA 113, 7035 (2016).DOI
  20. [20]S. Hong and D. Kim, “Medium-range order in amorphous ices revealed by persistent homology,” J. Phys. Condens. Mater. 31, 455403 (2019).DOI
  21. [21]J. R. Stellhorn, B. Paulus, S. Hosokawa, W.-C. Pilgrim, N. Boudet, N. Blanch, H. Ikemoto, S. Kohara, and Y. Sutou, “Structure of amorphous Cu₂GeTe₃ and the implications for its phase-change properties,” Phys. Rev. B 101, 214110 (2020).DOI
  22. [22]D. O. Morley, P. S. Salmon, and M. Wilson, “Persistent homology in two-dimensional atomic networks,” J. Chem. Phys. 154, 124109 (2021).DOI
  23. [23]A. Hirata, K. Matsue, and M. Chen, Structural analysis of metallic glasses with computational homology. (Springer, 2016).DOI
  24. [24]Y. Onodera, Y. Takimoto, H. Hijiya, T. Taniguchi, S. Urata, S. Inaba, S. Fujita, I. Obayashi, Y. Hiraoka, and S. Kohara, “Origin of the mixed alkali effect in silicate glass.” NPG Asia Mater. 11, 75 (2019).DOI
  25. [25]²⁵S. Hosokawa, J.-F. Bérar, N. Boudet, W.-C. Pilgrim, L. Pusztaí, S. Hiroi, K. Maruyama, S. Kohara, H. Kato, H. E. Fischer, and A. Zeidler, “Partial structure investigation of the traditional bulk metallic glass Pd₄₀Ni₄₀P₂₀,” Phys. Rev. B 100, 054204 (2019).
  26. [26]A. Hirata, T. Wada, I. Obayashi, and Y. Hiraoka, “Structural changes during glass formation extracted by computational homology with machine learning,” Commun. Mater. 1, 98 (2020).DOI
  27. [27]S. S. Sorensen, T. Du, C. A. Biscio, L. Fajstrup, and M. M. Smedskjaer, “Persistent homology: A tool to understand medium-range order glass structure,” Journal of Non-Crystalline Solids: X 16, 100123 (2022).DOI
  28. [28]Y. Lee, S. D. Barthel, P. Dotko, S. M. Moosavi, K. Hess, and B. Smit, “Quantifying similarity of pore-geometry in nanoporous materials,” Nat. Commun. 8, 15396 (2017).DOI
  29. [29]K. Sasaki, R. Okajima, and T. Yamashita, “Liquid structures characterized by a combination of the persistent homology analysis and molecular dynamics simulation,” AIP Conference Proceedings 2040, 020015 (2018).DOI
  30. [30]E. Minamitani, T. Shiga, M. Kashiwagi, and I. Obayashi, “Topological descriptor of thermal conductivity in amorphous Si,” J. Chem. Phys. 156, 244502 (2022).DOI
  31. [31]E. Minamitani, T. Shiga, M. Kashiwagi, and I. Obayashi, “Relationship between local coordinates and thermal conductivity in amorphous carbon,” J. Vacuum Sci. Technol. A 40, 033408 (2022).DOI
  32. [32]Z.-L. Wang, T. Ogawara, and Y. Adachi, “Property predictions for dual-phase steels using persistent homology and machine learning,” Adv. Theory Simul. 3, 1900227 (2020).DOI
  33. [33]Y. Jiang, D. Chen, X. Chen, T. Li, G.-W. Wei, and F. Pan, “Topological representations of crystalline compounds for the machine-learning prediction of materials properties,” npj Comput. Mater. 7, 28 (2021).DOI
  34. [34]A. Suzuki, M. Miyazawa, J. M. Minto, T. Tsuji, I. Obayashi, Y. Hiraoka, and T. Ito, “Flow estimation solely from image data through persistent homology analysis,” Sci. Rep. 11, 17948 (2021).DOI
  35. [35]S. Becker, E. Deijviver, R. Molinier, and N. Jakse, “Unsupervised topological learning approach of crystal nucleation,” Sci. Rep. 12, 3195 (2022).DOI
  36. [36]V. M. Nield, D. A. Keen, W. Hayes, and R. L. McGreevy, “Structure and fast-ion conduction in α-AgI,” Solid State Ionics 66, 247–258 (1993).DOI
  37. [37]S. Hull, “Superionics: crystal structures and conduction processes,” Rep. Prog. Phys. 67, 1233 (2004).DOI
  38. [38]F. Shimojo and M. Aniya, “Diffusion mechanism of Ag ions in superionic conductor Ag₂Se from ab initio molecular-dynamics simulations,” J. Phys. Soc. Jpn. 74, 1224 (2005).DOI
  39. [39]F. Shimojo, T. Inoue, M. Aniya, T. Sugahara, and Y. Miyata, “Ab Initio molecular-dynamics study of static structure and bonding properties of molten AgI,” J. Phys Soc. Jpn. 75, 114602 (2006).DOI
  40. [40]V. Bitrián and J. Trullàs, “Molecular dynamics study of polarization effects on AgI,” J. Phys. Chem. B 112, 1718 (2008).DOI
  41. [41]B. J. Morgan and P. A. Madden, “Relationships between atomic diffusion mechanisms and ensemble transport coefficients in crystalline polymorphs,” Phys. Rev. Lett. 112, 145901 (2014).DOI
  42. [42]O. Yamamoto, “Solid state ionics: a japan perspective,” Sci. Technol. Adv. Mater. 18, 504 (2017).DOI
  43. [43]T. M. Brenner, C. Gehrmann, R. Korobko, T. Livneh, D. A. Egger, and O. Yaffe, “Anharmonic host-lattice dynamics enable fast ion conduction in superionic AgI,” Phys. Rev. Mater. 4, 115402 (2020).DOI

Paper details

vlnc:W_9tv9j237
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents